Leveraged ETF Perps: Three Kinds of Leverage in One Position

What a leveraged ETF perp actually stacks

Several crypto perpetual venues now list markets named for US leveraged and inverse ETFs — SOXL, TQQQ, SOXS, TZA, TMF, TBT, UVXY and KORU among them. They look like any other perp ticker on the screen, and they are not.

Three separate mechanisms are at work in one position, and they are usually discussed as if they were one:

1. The fund's leverage. The ETF itself holds swaps and futures sized to deliver a stated multiple of its index — for one trading day. At each close it rebalances to restore that multiple against its new net assets.
2. The perpetual wrapper. The perp has no expiry and no delivery, so it is tethered to the fund's price by a funding payment exchanged between longs and shorts.
3. Your account leverage. Whatever margin multiple you choose on the venue applies on top of both of the above.

Only the first of these is unusual. Funding and margin behave the way they do on any perp, and if you are new to them, perpetual futures explained and how crypto leverage works cover the mechanics. The daily reset has no crypto-native analogue at all, and it is the layer that decides whether a correct view on semiconductors still loses money.

Why the multiple only holds for one day

A 3x fund does not promise you three times the index return over your holding period. It promises three times the index return today, and then it does the same thing again tomorrow from a different starting balance.

That resizing is the whole story. After a winning day the fund has more assets and buys more exposure; after a losing day it has fewer and sells. It is a mechanical buy-high, sell-low rebalance, executed once a day, and it is not optional — it is what keeps the multiple honest on a one-day view.

The consequence is that your return depends on the path the index took, not only on where it ended.

An index that goes nowhere, and two funds that both lose

Take a ten-day path with these daily index moves: +4.0%, −5.0%, +6.0%, −3.0%, −4.0%, +5.0%, +3.0%, −6.0%, +4.0%, −3.5%. Those are large days, but not implausible ones for a semiconductor index. The index finishes at 99.50 — down 0.50%, effectively flat.

Now apply a 3x long fund and a 3x inverse fund to the same path, starting both at 100. Day one, the index rises 4.0%, so the long fund gains 12.0% to 112.00 and the inverse fund loses 12.0% to 88.00. Day two, the index falls 5.0%: the long fund loses 15.0% of 112.00, not of its original 100, finishing at 95.20, while the inverse fund gains 15.0% of 88.00 to reach 101.20. Repeat that for eight more days.

Reading Figure 1

Both curves finish below where they started, and they were betting against each other. The long fund ends at 92.69 and the inverse fund at 89.91. There was no way to be on the right side of this, because there was no right side — the index barely moved, and both funds paid for the journey.

The loss is not a fee, and it is not tracking error. Three times a −0.50% index return is −1.50%. The long fund lost 7.31%, so 5.81 percentage points came from somewhere other than the index. It came from the resizing: every reversal caught the fund holding exposure sized for the move that just finished rather than the one about to start.

On day 8 the two curves touch, both at 92.47. That is not a plotting error. Over any pair of index moves the two funds apply the same two factors in the opposite order, and multiplication does not care about order — so on a symmetric enough stretch they land in exactly the same place. It is the clearest possible demonstration that this cost is structural rather than directional.

What to do with it: treat the holding period as part of the position, not a detail. If your view is "semis grind sideways for two weeks and then break out," a leveraged fund will charge you for the two weeks before the view pays. Size the entry for the path you actually expect, and if the honest answer is "I don't know the path," the leveraged wrapper is the wrong instrument for that view.

Compounding is not always a drag

The word "decay" gets attached to these products as though the arithmetic only ever runs one way. It does not.

Run the same 3x long fund along a smooth trend — ten consecutive days of +2.0% — and the index finishes at 121.90, up 21.90%. Three times that return would be +65.70%. The fund actually finishes at 179.08, up 79.08%: it beat its own multiple by 13.39 percentage points, because each day it reinvested the previous day's gain at three times leverage.

This is the same mechanism as before, with the sign flipped. Daily rebalancing adds to a trend and subtracts from a chop. The products are not designed to bleed; they are designed to be path-dependent, and path dependence pays when the path is straight.

The eight are not one product

The daily reset is common to all of these funds. Almost nothing else is. They get discussed as a single category — "the 3x ETFs" — when three of the eight are not 3x equity funds at all, and a fourth is a single-country bet.

Three of those rows deserve a second look before you size anything.

TMF and TBT are not a matched pair. They reference the same bond index, so they read like a bull/bear couple, but TMF is Direxion at +3x and TBT is ProShares at −2x. A "hedge" built by holding one against the other is lopsided by construction, and the arithmetic in Figure 1 applies to each of them at a different strength.

UVXY is the odd one out twice over. Its factor is 1.5x, not 3x, and its underlying is an index of short-term VIX futures rather than a stock index. Those futures carry their own roll cost that has nothing to do with the daily reset, so UVXY has two independent sources of path dependence stacked on each other. Nothing in this lesson's arithmetic captures the second one.

KORU runs on a different clock. It is a US-listed fund tracking Korean equities, and the Korean market closes hours before the US session even opens — so KORU's own US-hours price is already an estimate of a market that shut overnight. The perp then trades around the clock on top of that. Three timetables, one ticker, and only the middle one is the fund's.

What the perpetual wrapper adds

The perp changes two things about holding one of these funds, and leaves a third genuinely undocumented.

Funding is a two-way carry, not a built-in cost

It is tempting to describe funding as a third drag stacked under the expense ratio and the daily reset. That is wrong, and it matters.

On Aster, edgeX and Lighter alike, the funding rate is driven by the perp's premium or discount to its index price, with a small interest component and a clamp. Because the premium can be negative, the rate can be negative. edgeX's documentation states the convention directly: when the rate is positive longs pay shorts, and when it is negative shorts pay longs. Lighter's documentation says the same, and clamps the rate at 0.5% per hour; edgeX and Aster apply a ±0.05% dampener to the interest component of the formula.

So funding is a carry that depends on positioning, not a structural toll on longs. In practice a crowded long side in a popular leveraged ticker will often mean longs pay — but that is an observation about demand, not a property of the instrument, and it can reverse. The daily reset cannot.

The perp does not close when the exchange does

The underlying funds trade on NYSE Arca during US market hours. The perps trade continuously, which means price discovery carries on through nights, weekends and US holidays with no primary market to anchor it.

Aster documents what it does about this: during low-liquidity periods — overnight, weekends and public holidays — it applies an exponentially weighted moving average to the last traded price when forming the mark, with a smoothing factor of β = exp(−1/1600), and says the purpose is partly to avoid unnecessary forced liquidations in thin conditions. Funding continues to accrue across those hours on all three venues.

The practical consequence is that a fill taken at 3am on a Sunday is struck at a price the fund itself could not have traded, and a weekend gap gets absorbed into the mark rather than appearing as a Monday opening print.

What the perp marks against is not documented

Here the honest answer is that the venues do not say. Aster's documentation describes the mark-price construction and names Pyth among its index providers, but neither it nor edgeX's nor Lighter's documentation states whether an ETF perp's reference price is the fund's traded price on NYSE Arca, its once-daily net asset value, or something else assembled from the underlying index.

That gap matters less than it first appears, for one reason: the daily reset is embedded in the fund itself. Both its market price and its NAV already reflect every rebalance it has ever done. A market named for one share of SOXL inherits SOXL's path dependence whichever of the two it references — the only construction that would avoid it is one referencing the semiconductor index directly, and that would not be a SOXL market at all.

Can you harvest the decay by shorting both sides?

The natural trade to reach for, once you have seen Figure 1, is to short the long fund and the inverse fund at the same time and collect the shortfall from both. On the ten-day choppy path that works: shorting 100 of each returns 7.31 plus 10.09, or +8.70% on the 200 of notional committed.

Then run the same trade along the trends.

The two trend rows are mirror images, and that symmetry is the finding: the trade does not care which way the index goes, only whether it goes anywhere. Solve for the constant daily move at which the position breaks even and the answer is exactly zero. Any sustained drift, up or down, loses — the position is at its maximum when the index is perfectly flat and falls away on both sides.

So this is not a decay harvest and it is not market-neutral in any useful sense. It is a bet that realised movement stays choppy and directionless, with a payoff shaped like short-trend, long-chop. That is a legitimate position, but it is a view on the path, and it should be sized against the worst trend you think is plausible rather than against the index level.

Account leverage on top of fund leverage

The last layer is the one the venue hands you, and it multiplies rather than adds.

Read the last column as the actual risk statement. At 5x margin on a 3x fund, a 6.67% single-day move in the underlying index is enough to exhaust the margin — before fees, funding, or the maintenance buffer that triggers liquidation somewhat earlier than the arithmetic suggests.

The unlevered version of that arithmetic is in the prospectus. A 3x fund is wiped out by a 33% single-day fall in its index, because three times 33% is everything you put in — and Direxion states exactly that outcome in SOXL's summary prospectus. Every row of the table above is that same calculation with your own margin dividing the threshold down.

The practical error is anchoring on the venue's leverage number alone. "5x" reads as moderate next to the 20x and 50x available on major crypto pairs, and on a 3x fund it is not moderate — it is the equivalent of 15x on the index itself.

Risks and limits

Three further limits are worth stating plainly. The fund figures here exclude the expense ratio and the fund's own financing costs, so a real holding underperforms these numbers slightly. The perp adds funding on top, which the arithmetic above does not include. And a perp on an instrument whose underlying trades only during exchange hours can price away from that instrument while the exchange is closed, so an entry or exit taken outside those hours is not necessarily struck at a level the fund itself could have traded.

Conclusion

A leveraged ETF perp is three positions wearing one ticker. The fund's daily reset makes the return path-dependent, which is a genuine cost in a chop and a genuine benefit in a trend — Figure 1 shows a flat index costing a long 7.31% and an inverse 10.09% at the same time, while a smooth 21.90% rally hands the long 79.08% against a 65.70% benchmark. The perpetual wrapper adds funding and removes expiry. Your own margin multiplies whatever the first two produce, so a 5x account setting is 15x index exposure and roughly a 6.67% single-day index move from a wipeout.

None of that makes the instrument unusable. It makes the forecast you need more specific than the one most people think they are making: not a direction, but a direction and a path, held over a stated number of days. If you cannot state the path, prefer the unleveraged perp on the same index and take the leverage — if you want it — as an account setting you can see and adjust, rather than one compounded into the instrument itself.

Frequently asked questions